Towards Ubiquitous Personalized Music Recommendation with Smart Bracelets

Nowadays, recommender systems play an increasingly important role in the music scenario. Generally, music preferences are related to internal and external conditions. For example, mood state and ongoing activity will affect users' music preferences. However, conventional music recommenders cannot capture these conditions since they only utilize the online data but ignore the impact of physical-world information. In this paper, we leverage the contexts from low-cost smart bracelets for ubiquitous personalized recommendation to meet users' music preference. We first conduct a large-scale questionnaire survey, which illustrates moods, activities, and environments will affect music preferences. Then we perform a one-week field study among 30 participants, where they receive personalized music recommendation and record preferences and mood. Meanwhile, participants' context information is collected with bracelets. Analyses on the data demonstrate significant relationships between music preference, mood, and bracelet contexts. Furthermore, we propose a novel Multi-task Ubiquitous Music Recommendation model (MUMR) to predict personalized music preference with bracelet contexts as input and mood prediction as an auxiliary task. Experiments show significant improvement in music recommendation performances with MUMR. Our work demonstrates the possibility of ubiquitous personalized music recommendations with smart bracelets data, which is an encouraging step towards building recommender systems aware of physical-world contexts.